We study the automatic identification of educational items worthy of content authors' attention. Based on the results of such analysis, content authors can revise and improve the content of learning environments. We provide an overview of item properties relevant to this task, including difficulty and complexity measures, item discrimination, and various forms of content representation. We analyze the potential usefulness of these properties using both simulation and analysis of real data from a large-scale learning environment. We also describe two case studies where we practically apply the identification of attention-worthy items. Based on the analysis and case studies, we provide recommendations for practice and impulses for further research.